🧠 Brain Tumor Analysis System

A sophisticated web-based application for brain tumor classification using deep learning. This project leverages transfer learning with ResNet50 to classify MRI brain images into four categories: No Tumor, Meningioma, Glioma, and Pituitary tumors.

📋 Table of Contents

• Features
• Technology Stack
• Project Structure
• Installation
• Usage
• Model Architecture
• Dataset Information
• Important Notes
• Future Improvements
• License

---

✨ Features

• Web-based Interface: Modern, responsive UI with gradient animations and professional design
• Real-time Image Classification: Instant brain tumor classification from uploaded MRI scans
• Transfer Learning: Powered by ResNet50 architecture with pretrained ImageNet weights
• Four-class Classification: Distinguishes between:
  • No Tumor
  • Meningioma
  • Glioma
  • Pituitary tumors
• User-Friendly Design: Intuitive file upload with visual feedback and clear results display
• Mobile Responsive: Fully responsive design for desktop, tablet, and mobile devices
• Professional UI/UX: Modern gradient animations, smooth transitions, and polished visual elements

---

🛠 Technology Stack

Component	Technology
Backend Framework	Flask 3.1.2
Deep Learning	PyTorch 2.10.0
Computer Vision	torchvision 0.25.0
Image Processing	Pillow 12.1.0
Numerical Computing	NumPy 2.4.1
Frontend	HTML5, CSS3, Vanilla JavaScript
Python Version	3.8+

---

📁 Project Structure

Brain Tumor Analysis/
├── app.py                              # Flask application main file
├── initialize_model.py                 # Script to create untrained model
├── requirements.txt                    # Python dependencies
├── README.md                          # Project documentation
├── models/
│   └── bt_resnet50_model.pt           # Trained/untrained model weights
├── static/
│   ├── css/
│   │   └── main.css                   # Styled CSS with animations
│   ├── images/
│   │   ├── brain_logo.png             # Project logo
│   │   └── (uploaded images)          # User uploaded MRI scans
├── templates/
│   ├── index.html                     # Home page with upload form
│   └── pred.html                      # Results page with classification
├── test_images/                       # Sample test images
├── venv/                              # Python virtual environment
└── brain_tumor_dataset_preparation.ipynb  # Dataset preparation notebook

---

🚀 Installation

Prerequisites

• Python 3.8 or higher
• pip (Python package manager)
• 2GB RAM minimum
• CUDA 12.x (optional, for GPU acceleration)

Step 1: Clone or Navigate to Project

cd "Brain Tumor Analysis"

Step 2: Create Virtual Environment

# Create virtual environment
python3 -m venv venv

# Activate virtual environment
# On Linux/Mac:
source venv/bin/activate

# On Windows:
venv\Scripts\activate

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Initialize Model (First Time Only)

python initialize_model.py

This creates the untrained model architecture saved to models/bt_resnet50_model.pt.

---

💻 Usage

Starting the Application

# Activate virtual environment (if not already activated)
source venv/bin/activate

# Run Flask development server
python app.py

The application will start on http://127.0.0.1:5000

Using the Web Interface

1. Open Browser: Navigate to http://127.0.0.1:5000
2. Upload Image: Click the upload area or drag-and-drop a brain MRI scan (JPG, PNG, etc.)
3. Analyze: Click "Analyze Image" button
4. View Results: See classification result with color-coded output:
   • 🟢 Green: No Tumor
   • 🟠 Orange: Meningioma
   • 🔴 Red: Glioma
   • 🟣 Purple: Pituitary tumor
5. Analyze Another: Click button to return to home and analyze another image

---

🧠 Model Architecture

ResNet50 Backbone

• Base Model: ResNet50 with ImageNet pretrained weights
• Input Size: 512×512 RGB images
• Preprocessing:
  • Resize to 512×512
  • Normalize with ImageNet mean/std
  • Convert to tensor

Classification Head

• Layer 1: 2048 → 512 (ReLU activation)
• Dropout: 0.5 (prevents overfitting)
• Layer 2: 512 → 256 (ReLU activation)
• Dropout: 0.3
• Layer 3: 256 → 128 (ReLU activation)
• Output Layer: 128 → 4 (softmax for multi-class classification)

Model Summary

Total Parameters: ~25.5 Million
Trainable Parameters: ~3.3 Million (classification head)
Frozen Parameters: ~23.5 Million (ResNet50 backbone)

---

📊 Dataset Information

Brain Tumor Dataset

The model is designed to work with the Brain Tumor MRI Dataset from Figshare:

• Source: https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset
• Classes: 4 types (No Tumor, Meningioma, Glioma, Pituitary)
• Image Count: ~7,000 total images
• Image Format: JPG/PNG
• Image Dimensions: Variable (preprocessing handles resizing)

Training Recommendations

- Training-Validation Split: 70-30
- Batch Size: 32-64
- Learning Rate: 0.0001 (Adam optimizer)
- Epochs: 50-100
- Loss Function: CrossEntropyLoss
- Early Stopping: Monitor validation accuracy

---

⚠️ Important Notes

Current Model Status

⚠️ The model shipped with this project is NOT trained

• The initialize_model.py script creates an untrained ResNet50 architecture
• Random predictions are returned until proper training is performed
• For resume/production use, you must train the model on actual brain tumor data

Training the Model

To use this project effectively:

1. Obtain Dataset: Download from Kaggle or Figshare
2. Prepare Data: Use brain_tumor_dataset_preparation.ipynb notebook
3. Train Model:
   • Use PyTorch training loop
   • Modify app.py to load trained weights
   • Test on validation set (aim for 90%+ accuracy)
4. Evaluate: Test with medical MRI samples
5. Deploy: Update model path in Flask app

Disclaimer

This tool is experimental and should NOT be used for:

• Actual medical diagnosis
• Clinical decision making
• Patient treatment planning

Always consult qualified medical professionals for brain tumor diagnosis and treatment.

---

🔧 Configuration

Modify Classification Labels

Edit in app.py line ~45:

class_names = ['None', 'Meningioma', 'Glioma', 'Pituitary']

Change Upload Directory

Edit in app.py line ~30:

UPLOAD_FOLDER = 'static/images'

Adjust Image Size

Edit in app.py line ~35:

IMG_SIZE = 512  # Change as needed

---

📝 API Endpoints

GET /

• Description: Render home page
• Response: HTML page with upload form

POST /

• Description: Upload image and get classification
• Parameters: Form data with bt_image file
• Response: HTML results page with prediction

POST /predict

• Description: API endpoint for JSON predictions
• Request: JSON with base64 encoded image
• Response: JSON with classification results

---

🎨 UI/UX Features

Design Highlights

• Color Scheme: Professional navy blue (#0f172a) and medical teal (#0891b2)
• Animations: Smooth fade-in, slide-up, and pulse animations
• Responsive Design: Works on 320px to 4K+ screens
• Accessibility: High contrast ratios and readable typography
• Interactive Feedback: Hover effects on buttons and cards

CSS Variables

Customize colors in static/css/main.css:

--primary: #0f172a;           /* Navy blue */
--secondary: #0891b2;          /* Medical teal */
--accent-success: #059669;     /* Green */
--accent-warning: #d97706;     /* Orange */
--accent-danger: #dc2626;      /* Red */

---

🚀 Future Improvements

Phase 1: Model Enhancement

• Train ResNet50 on brain tumor dataset
• Implement class imbalance handling
• Add confidence scores to predictions
• Create model evaluation metrics dashboard

Phase 2: Feature Expansion

• Image preprocessing filters (brightness, contrast adjustment)
• Batch image upload and analysis
• Export results to PDF/CSV
• User authentication and history
• Dark mode toggle

Phase 3: Advanced Features

• Explainable AI (Grad-CAM visualization)
• Ensemble models for better accuracy
• DICOM file support
• Real-time prediction API
• Docker containerization

Phase 4: Production Ready

• Unit and integration testing
• Error handling and logging
• Load balancing for multiple users
• Database for predictions history
• API rate limiting
• Deployment to cloud (AWS/GCP/Azure)

---

📖 File Descriptions

File	Purpose
app.py	Main Flask application with routes and model inference
initialize_model.py	Creates untrained model weights for development
brain_tumor_dataset_preparation.ipynb	Jupyter notebook for dataset preparation and exploration
requirements.txt	Python package dependencies
templates/index.html	Home page template with upload form
templates/pred.html	Results page template showing predictions
static/css/main.css	Comprehensive styling with animations and gradients
static/images/	Directory for uploaded and logo images

---

🔍 Troubleshooting

Issue: Model file not found

Solution: Run python initialize_model.py to create model weights

Issue: Port 5000 already in use

# Kill process using port 5000
lsof -ti:5000 | xargs kill -9

# Or run on different port
python -c "from app import app; app.run(port=5001)"

Issue: CUDA out of memory

# Use CPU instead in app.py
device = torch.device('cpu')

Issue: Image upload not working

Solution: Ensure templates/ folder exists and check Flask upload folder permissions

---

📚 References

• PyTorch Documentation
• Flask Documentation
• ResNet Paper
• Brain Tumor Dataset
• Transfer Learning Guide

---

📄 License

This project is provided as-is for educational and portfolio purposes.

Medical Disclaimer: This tool is not approved for medical diagnosis. Always consult qualified medical professionals for any health-related decisions.

---

👤 Author

Adithya
Portfolio Project: Brain Tumor Analysis System

---

💬 Key Skills Demonstrated

• ✅ Deep Learning (PyTorch, Transfer Learning)
• ✅ Full-Stack Web Development (Flask, HTML/CSS/JS)
• ✅ Computer Vision (ResNet50, Image Classification)
• ✅ UI/UX Design (Responsive, Modern CSS)
• ✅ Project Structure & Documentation
• ✅ Version Control & Best Practices